Eragrostis (Poaceae): Monophyly and Infrageneric Classification Amanda L

Home , Calamovilfa gigantea, Cladoraphis, Coelachyrum, Eragrostis, Pappophorum bicolor, Pogonarthria

Aliso: A Journal of Systematic and Evolutionary Botany

Volume 23 | Issue 1 Article 44

2007 Eragrostis (Poaceae): Monophyly and Infrageneric Classification Amanda L. Ingram Cornell University, Ithaca, New York

Jeff .J Doyle Cornell University, Ithaca, New York

Follow this and additional works at: http://scholarship.claremont.edu/aliso Part of the Ecology and Evolutionary Biology Commons, and the Plant Sciences Commons

Recommended Citation Ingram, Amanda L. and Doyle, Jeff .J (2007) "Eragrostis (Poaceae): Monophyly and Infrageneric Classification," Aliso: A Journal of Systematic and Evolutionary Botany: Vol. 23: Iss. 1, Article 44. Available at: http://scholarship.claremont.edu/aliso/vol23/iss1/44 Aliso 23, pp. 595–604 ᭧ 2007, Rancho Santa Ana Botanic Garden

ERAGROSTIS (POACEAE): MONOPHYLY AND INFRAGENERIC CLASSIFICATION

AMANDA L. INGRAM1,2 AND JEFF J. DOYLE L. H. Bailey Hortorium, Cornell University, 228 Plant Science, Ithaca, New York 14853, USA 1Corresponding author ([email protected])

ABSTRACT Eragrostis is a large genus in subfamily Chloridoideae of Poaceae. Recent phylogenetic analyses have suggested that the genus may not be monophyletic, that some of its segregate genera may be better placed within Eragrostis, and that current infrageneric classiﬁcations may not represent monophyletic groups. We have used molecular sequence data from the plastid locus rps16 and the nuclear gene waxy from a broad sample of Eragrostis species and representatives of six of the seven segregate genera to address these issues. We found that Eragrostis is monophyletic with the inclusion of several of the segregates, including Acamptoclados, Diandrochloa, and Neeragrostis. The placement of Cla- doraphis and Stiburus is uncertain. Thellungia does not belong in Eragrostis and is actually more closely related to Sporobolus. These data also suggest that existing infrageneric classiﬁcations are inadequate and do not correspond to monophyletic groups within Eragrostis. Key words: Acamptoclados, Cladoraphis, Diandrochloa, Eragrostis, Neeragrostis, Pogonarthria, rps16, Stiburus, Thellungia, waxy.

INTRODUCTION genus, see Van den Borre and Watson 1994). Since the original description there has been little agreement as to which Eragrostis Wolf is a genus of approximately 350 species species actually belong in the genus and how they are related in Poaceae (Watson and Dallwitz 1992). It is the largest ge- to each other. Species in Acamptoclados Nash, Cladoraphis nus in subfamily Chloridoideae, a group comprising about Franch., Diandrochloa De Winter, Eragrostiella Bor, Neer- 1500 species (Van den Borre and Watson 1997). Members agrostis Stiburus Thellungia of Eragrostis generally are characterized by paniculate inflo- Bush, Stapf, and Stapf have rescences, multi-floreted spikelets, glabrous three-nerved been included in Eragrostis at various times. Clayton and Renvoize (1986) also suggested a close relationship between lemmas, ciliate ligules, and C4 photosynthesis. The genus is morphologically and anatomically diverse, however, and ex- Eragrostis and Pogonarthria Stapf. hibits a wide range of variation in many characters. For in- The various modes of spikelet disarticulation that are seen stance, the panicles range from very loose and open to high- in the genus have been the most common source of char- ly contracted spicate structures (Watson and Dallwitz 1992). acters for delimiting infrageneric groups. Spikelets may dis- NAD-ME, PCK-like, and intermediate forms of leaf blade articulate from the top, from the bottom, or as a unit. Ad- anatomy are found in the genus (Van den Borre and Watson ditionally, whether the paleas are retained on the rachilla or 1994). Several major types of the bicellular microhairs com- fall with the lemmas can be an important character, as can mon to Chloridoideae are found in Eragrostis, including the the persistence of the rachilla. These characters are quite chloridoid type, the panicoid type, the Pappophorum type, variable in Eragrostis and can be seen in a number of dif- and intermediates (Prendergast et al. 1986; Amarasinghe and ferent combinations. Unfortunately, these characters are not Watson 1990). Eragrostis species range throughout the generally as useful as one might hope. There may be tem- world’s tropical and subtropical regions, and they are most poral variation that is not always obvious at particular stages commonly found in weedy disturbed areas and in dry hab- in the life cycle (e.g., the paleas may be retained slightly itats. Most of the species are of little economic importance, longer than the lemmas but are still deciduous). From a prac- but one species (E. tef; tef) is cultivated as a major cereal tical standpoint, this means that herbarium material may be crop in Ethiopia. This species is also an important forage impossible to score or misleading if not collected at precisely grass, as are several other species, including E. cilianensis the correct stage. These intermediacies also cloud the dis- and E. curvula. Due in part to its large size and wide geo- tinction of the character states. The infrageneric classifica- graphic distribution, there has been no comprehensive tax- tions proposed by Koch (1848) and numerous later botanists onomic treatment of Eragrostis, and there has been some relied heavily on these characters (see Van den Borre and debate in the recent literature as to whether the genus is Watson 1994). More recent classifications have included monophyletic (Van den Borre and Watson 1997; Hilu and some other morphological characters, such as spikelet shape Alice 2001) and how infrageneric groups should be delim- (Lazarides 1997), pubescence on the palea keels, panicle ited (reviewed in Van den Borre and Watson 1994). branching, lemma keel and margin shape and curvature, and Eragrostis was first described by Wolf (1776) from ma- floret fertility (Cope 1998). terial of E. minor (for a detailed taxonomic history of the A recent phylogenetic analysis of anatomical and morphological data from 56 species of Eragrostis and two seg- 2Present address: Wabash College, PO Box 352, Crawfordsville, regate genera by Van den Borre and Watson (1994) led the Indiana 47933, USA. authors to conclude that divisions based on these spikelet 596 Ingram and Doyle ALISO disarticulation characters did not represent natural groups. doraphis is a genus of two species distributed in southern They instead suggested that the genus could be divided into Africa. The plants are halophytic or glycophytic and have two subgenera based on a number of correlated anatomical free pericarps and a unique ‘‘armed’’ growth form. Panicle and morphological characters. However, no uncontradicted branches in Eragrostis species terminate with spikelets, but synapomorphies were identified for these groups, and there the terminal spikelets in Cladoraphis are lacking. This in are numerous exceptions to most of the characters proposed combination with hardened panicle branches and tough, in- as diagnostic of the subgenera. The results from the phylo- rolled leaves produces thorny appendages. Diandrochloa, a genetic analyses in their study may also have been affected genus of seven widely distributed species, was separated by the frequent occurrence of allopolyploidy in the genus. from Eragrostis by De Winter (1960) on the basis of mem- Approximately 69% of the species in the genus are poly- branous ligules (other Eragrostis species have a ligule that ploids (Hunziker and Stebbins 1986), and many of the taxa is a fringe of hairs; Watson and Dallwitz 1992). Phillips included in the Van den Borre and Watson (1994) study are (1982) rejected this treatment, however, and included these known allopolyploids. Taxa of hybrid origin can exhibit species in Eragrostis. Eragrostiella comprises five species unique combinations of morphological and anatomical char- in southern Asia and Australia and was separated from Er- acter states and when included in phylogenetic analyses can agrostis by Bor (1940) based on the spicate inflorescences produce misleading hypotheses of relationships (McDade and tufted habit found in these species. Neeragrostis was 1990, 1992). described by Bush (1903) based on a single dioecious spe- The monophyly of Eragrostis has been brought into ques- cies (ϭ N. reptans). Two additional species of Eragrostis, tion by recent phylogenetic analyses of subfamily Chlori- one dioecious and another with hermaphrodite flowers (E. doideae. Van den Borre and Watson (1997) analyzed a data hypnoides), have been positioned in Neeragrostis by some set of anatomical and morphological characters from all 166 authors. All three species are prostrate annuals distributed in chloridoid genera (as recognized by Watson and Dallwitz the Americas. Stiburus, a small genus of two species from 1992). The authors used the two subgenera of Eragrostis southern Africa, was first described by Stapf (1900). These identified in their 1994 analysis of the genus as terminals species were separated from Eragrostis on the basis of their for their phylogenetic analyses and found that they were not villous spikelets, but Phillips (1982) pointed out that a num- sister taxa and were in fact found in two widely divergent ber of Old World Eragrostis species have pubescent spike- clades. Hilu and Alice (2001) conducted a phylogenetic anal- lets and submerged Stiburus into Eragrostis. The monotypic ysis of sequence data from the plastid locus matK from 74 genus Thellungia (ϭ E. advena) is native to Australia and species in 56 chloridoid genera. Included in the sample were was first described in 1920 by Stapf from alien material seven Eragrostis species, and the authors also found that found growing among wool refuse near a worsted mill in these species did not form a monophyletic group. Five of Switzerland (Stapf 1920). Thellungia advena was originally the species grouped together in a clade sister to Eragros- thought to have close affinities to Sporobolus R. Br. due to tiella, a genus thought to be closely related to Eragrostis, its similar floral structure and free pericarps, but the author but two other species were found in a different clade. Even more surprisingly, Pappophorum bicolor was embedded noted that T. advena differed from Sporobolus in having within the larger Eragrostis clade. This result was unex- multiple florets per spikelet (as is found in Eragrostis). Phil- pected because Pappophorum Schreb. bears little morpho- lips (1982) placed this species in Eragrostis, pointing out logical similarity to Eragrostis, and this relationship had that the floral structure of T. advena fit her concept of Er- never been suggested before. agrostis and that free pericarps are found elsewhere in the Although these studies have raised the important issue of genus. The Van den Borre and Watson (1997) analyses sug- whether Eragrostis is a natural group, they suffered from gested that Acamptoclados, Diandrochloa, Neeragrostis, and some limitations. The Van den Borre and Watson (1997) Thellungia are closely related to Eragrostis, but because study used composite terminals. These groups are far from composite terminals were used, it is impossible to determine uniform in many of the characters used in the phylogenetic whether these genera are separate lineages or if they are analysis, which necessitates coding a number of polymor- better included within Eragrostis. Hilu and Alice (2001) phisms in the matrix. In addition, the monophyly of these sampled Eragrostis advena (ϭ Thellungia advena) and two subgenera has not been confirmed by phylogenetic anal- found in their analysis that it did not group with the clade yses of independent data sets (Ingram and Doyle 2003). Both containing most of the other Eragrostis species and Er- of these factors could have seriously affected the outcome agrostiella. of the phylogenetic analysis of the subfamily. The Hilu and Here we assess the monophyly of Eragrostis and the re- Alice (2001) study suffered from poor sampling of Era- lationships of the segregate genera by conducting phyloge- grostis species and the probable misidentification of some netic analyses of a broad sample of Eragrostis species, in- plant materials (see Results). cluding taxa from both subgenera identified by Van den Another complicating factor in determining whether Er- Borre and Watson (1994) and representatives of the major agrostis is a monophyletic group is the profusion of segre- spikelet disarticulation types and other infrageneric taxa that gate genera. Acamptoclados, a monotypic genus (ϭ E. ses- have been recognized in the genus. We include all available silispica) from the plains and prairies of North America, was segregate genera and other genera identified by previous separated from Eragrostis by Nash on the basis of its free phylogenetic analyses to be potentially closely related to Er- pericarp and lack of secondary inflorescence branches agrostis. The fit of the relationships uncovered by these mo- (Small 1903). Phillips (1982) treated Acamptoclados as a lecular data to existing infrageneric classification systems is synonym of Eragrostis in her analysis of Eragrostideae. Cla- also examined. We use sequence data from the plastid locus VOLUME 23 Eragrostis Classification 597 rps16 and a portion of the nuclear gene waxy (granule-bound RESULTS starch synthase I; GBSSI) in our phylogenetic analyses. Some accessions obtained from the USDA were misidentified in the NPGS database (Germplasm Resources Infor- MATERIALS AND METHODS mation Network). These accessions are marked with an as- Sampling terisk in Table 1. Also, the accession identified as Pappo- phorum bicolor turned out to be a mixed collection. This is Thirty-seven Eragrostis species were sampled, represent- the same accession whose matK sequence in the Hilu and ing all of the major morphological groups in the genus, Alice (2001) analysis of Chloridoideae was embedded within based primarily on spikelet disarticulation type (Table 1). a clade of Eragrostis sequences and suggested that Pappo- Representatives from both subgenera as delimited by Van phorum was nested within Eragrostis. When grown in the den Borre and Watson (1994) were included. Species re- Cornell University greenhouse, this seed lot was found to presenting six of the seven segregate genera were also in- contain a mixture of seeds from E. botryodes and P. bicolor, cluded in the taxon sample (material of Eragrostiella was so it appears that the matK sequence obtained from this ac- not available). Additionally, 18 species from other chloridoid cession was actually from an E. botryodes individual. Se- genera that are potentially close relatives of Eragrostis were quences from rps16 and waxy for P. bicolor resolve as sister included to test the monophyly of Eragrostis. Plant materials to P. mucronulatum in our analyses, and both of these spe- were obtained from the USDA National Plant Germplasm cies fall well outside Eragrostis (Fig. 1, 2). System (NPGS), herbarium specimens, or from personal col- Hilu and Alice (2001) also showed that Eragrostiella bra- lections (Table 1). Voucher specimens are deposited at Cor- chyphylla (Stapf) Bor was sister to a core Eragrostis clade. nell University (BH) except where noted in Table 1. Sequences for waxy and rps16 that we obtained from the same DNA used in that study are identical to sequences ob- DNA Extraction, Sequencing, and Data Analysis tained from Eragrostis tef, suggesting that this material may also have been misidentified. Voucher specimens were not DNA was isolated from fresh greenhouse-grown, field- made available to confirm this suspicion, but given that Er- collected and silica-dried, or herbarium material using the agrostiella brachyphylla bears little morphological similarity DNeasy kit (QIAGEN, Valencia, California, USA) according to Eragrostis tef, it seems unlikely that their DNA sequences to manufacturer’s instructions. PCR amplification, sequenc- would be identical. Additionally, seed identified as Eragros- ing, and sequence alignment of waxy and rps16 were per- tiella brachyphylla that we requested from the USDA turned formed as described in Ingram and Doyle (2003). Some of out to be tef. The Hilu and Alice (2001) accession was re- the sequences used in this study were previously published ported to be from a different source, but the coincidence is (Ingram and Doyle 2003: GenBank accessions AY136828– striking. AY136942); new sequences were deposited in GenBank (ac- A number of species (including Cladoraphis spinosa, cession numbers AY508649–AY508691, AY509525, and Neeragrostis reptans, Stiburus conrathii, and Uniola pittieri) AY509526). All sequences were aligned with Clustal࿞X were included in the rps16 data set but not in the waxy data (Thompson et al. 1997) using a variety of gap opening and set. The DNAs for these taxa were extracted from old, poor- extension penalties to test for the sensitivity of the phylo- ly preserved herbarium material and were so degraded that genetic analyses to different alignments. The aligned se- it was impossible to amplify the nuclear gene. quences were read into WinClada vers. 1.00.08 (Nixon The rps16 sequences were easily aligned, and the phylo- 1999a) for manual alignment adjustment, phylogenetic ana- genetic analysis recovered 351 equally parsimonious trees lysis, and tree manipulation. Unequivocal gaps (excluding (length [L] ϭ 230, consistency index [CI] ϭ 72, retention autapomorphies) were coded as presence/absence characters index [RI] ϭ 88). The strict consensus tree (Fig. 1) was following the simple gap coding method of Simmons and poorly resolved within Eragrostis, but this data set was able Ochoterena (2000). Coelachyrum piercei, a chloridoid grass to resolve relationships among the other chloridoid genera not thought to be included in Eragrostis, was used to root included in the analysis. The bootstrap values (Fig. 1) were the phylogenetic trees. This species was chosen on the basis generally quite high and showed high levels of support for of its early divergence in pilot analyses that included rps16 many of the major clades. However, support for clades with- and waxy sequences from Zea mays L., a panicoid grass. in Eragrostis was generally low. The data sets were initially analyzed separately due to the The waxy sequences were difficult to align, particularly in presence of allopolyploid species in the taxon sample. To the rapidly evolving introns in taxa outside Eragrostis. The assess the fit of existing infrageneric classification systems exons, however, were simple to align across all of the taxa. to the relationships defined by these molecular data, the Three equally parsimonious trees (L ϭ 4995, CI ϭ 40, RI polyploid sequences were culled from both the waxy and the ϭ 62) were recovered from the phylogenetic analysis of the rps16 data sets. These smaller sets of sequences were then complete data set. The strict consensus tree (Fig. 2) was combined into a single matrix and analyzed simultaneously. much better resolved within Eragrostis than with the rps16 All parsimony analyses were conducted with WinClada and data set. Bootstrap values were generally lower for this data NONA (Goloboff 1993), using traditional heuristic search set (Fig. 2), but the Eragrostis clade was very strongly sup- strategies and the ratchet (Nixon 1999b). Strict consensus ported. trees were calculated in WinClada. Bootstrap analyses were The rps16 and waxy data sets were analyzed both with performed in WinClada to assign support values to the and without (data not shown) the gap characters. The topol- clades. ogy of the trees resulting from both types of analyses were 598 Ingram and Doyle ALISO

Table 1. Collection data for taxa included in this study. USDA accessions (indicated by ‘‘PI’’ or ‘‘NSL’’) marked with an asterisk (*) were misidentiﬁed in the Germplasm Resources Information Network. Vouchers not located at Cornell University (BH) are indicated with the herbarium abbreviation in parentheses next to the voucher number.

Taxon Collection Locality Voucher

Acamptoclados sessilispicus (Buckley) Nash Texas, USA Kruse 256 (TAES) Calamovilfa gigantea (Nutt.) Scribn. & Merr. NSL 22960 New Mexico, USA Ingram 01-02 C. longifolia (Hook.) Hack. ex Scribn. & Southw. PI 433949 Missississpi, USA Ingram 02-02 Cladoraphis spinosa (L. f.) S. M. Phillips Braun 5732 South Africa Ingram 15-02 Coelachyrum piercei (Benth.) Bor PI 197534 Ethiopia Ingram 04-02 Dactyloctenium aegyptium (L.) Willd. PI 215592 Punjab, India Ingram 05-02 D. australe Steud. PI 299588 Cape Province, South Africa Ingram 06-02 D. giganteum B. S. Fisher & Schweick. PI 364504 Natal, South Africa Ingram 07-02 D. radulans (R. Br.) P. Beauv. PI 238276 Queensland, Australia Ingram 08-02 Diandrochloa japonica (Thunb.) A. N. Henry PI 213410 India Ingram 33-01 Eleusine coracana (L.) Gaertn. PI 462423 Bihar State, India Ingram 12-02 Enneapogon scoparius Stapf PI 208126 Transvaal, South Africa Ingram 03-02 Eragrostis airoides Nees PI 309995 Brazil Ingram 14-02 E. aspera (Jacq.) Nees PI 368248 Zimbabwe Ingram 01-99 E. bahiensis Schrad. ex Schult. PI 204185 Uruguay Ingram 01-01 E. barrelieri Daveau Arizona, USA Reeder 9835 E. bicolor Nees PI 165732 South Africa Ingram 02-99 E. botryodes Clayton 2000-09 Chafanna, Ethiopia Ingram 02-01 2000-13 Debre Birhan, Ethiopia Ingram 03-01 E. chapelieri (Kunth) Nees 2000-17 Ziha, Ethiopia Ingram 06-01 E. cilianensis (Bellardi) Vignolo ex Janch. PI 299912 South Africa Ingram 03-99 2000-24 Tis Abay, Ethiopia Ingram 07-01 E. ciliaris (L.) R. Br. Florida, USA Lewis 050-01 E. curvula (Schrad.) Nees PI 226071 Kenya Ingram 04-99 E. dielsii Pilg. ex Diels & Pritz. PI 238301 Australia Ingram 08-01 E. echinochloidea Stapf PI 184741 South Africa Ingram 09-01 E. elegantissima Chiov. 2000-16 Ziha, Ethiopia Ingram 10-01 E. heteromera Stapf PI 208129 South Africa Ingram 12-01 E. hypnoides (Lam.) Britton, Sterns & Poggenb. Mississippi, USA Alford 2829 E. intermedia Hitchc. PI 216400 Mexico Ingram 13-01 E. lehmanniana Nees PI 226073 Kenya Ingram 15-01 E. lugens Nees PI 203862 Brazil Ingram 16-01 E. macilenta (A. Rich.) Steud. PI 194929 Ethiopia Ingram 05-99 E. mexicana (Hornem.) Link PI 203652 Brazil Ingram 06-99 E. minor Host PI 223367 Iran Ingram 16-99 E. neesii Trin. PI 203650 Brazil Ingram 18-01 E. nutans (Retz.) Nees ex Steud. PI 217616 India Ingram 19-01 E. paniciformis (A. Braun) Steud. 2000-03 Wolaita Sodo, Ethiopia Ingram 20-01 E. papposa (Roem. & Schult.) Steud. 2000-01 Awasa, Ethiopia Ingram 22-01 E. patenti-pilosa Hack. 2000-26 Tis Abay, Ethiopia Ingram 23-01 E. pilosa (L.) P. Beauv. PI 213255* India Ingram 07-99 PI 219588 Pakistan Ingram 08-99 PI 221926 Afghanistan Ingram 09-99 PI 222988 Iran Ingram 10-99 E. polytricha Nees PI 202443 Chile Ingram 24-01 E. rigidior Pilg. 2000-07 Gidole, Ethiopia Ingram 26-01 E. schweinfurthii Chiov. 2000-12 Ametsegna Ager, Ethiopia Ingram 28-01 E. secundiﬂora J. Presl PI 216405 Texas, USA Ingram 27-01 E. tef (Zucc.) Trotter ‘Red Dabi’ PI 557457 Ethiopia Ingram 12-99 E. tenella (L.) P. Beauv. ex Roem. & Schult. PI 320980* Sierra Leone Ingram 32-01 E. tremula Hochst. ex Steud. PI 220220 Liberia Ingram 30-01 E. trichophora Coss. & Durieu PI 364802 South Africa Ingram 17-99 E. truncata Hack. PI 299962 South Africa Ingram 31-01 E. unioloides (Retz.) Nees ex Steud. PI 213254 India Ingram 16-02 Fingerhuthia sesleriiformis Nees PI 299968 South Africa Ingram 11-02 Leptochloa dubia (Kunth) Nees PI 216460 Mexico Ingram 14-99 Neeragrostis reptans (Michx.) Nicora Texas, USA Kruse 284 Pappophorum bicolor E. Fourn. PI 216526 Mexico Ingram 09-02 P. mucronulatum Nees PI 477097 Uruguay Ingram 10-02 Pogonarthria squarrosa (Roem. & Schult.) Pilg. Mpumalanga, South Africa Snow 7023 (MO) Schmidtia pappophoroides Steud. ex J. A. Schmidt PI 209163 South Africa Ingram 17-02 VOLUME 23 Eragrostis Classiﬁcation 599

Table 1. Continued.

Taxon Collection Locality Voucher

Spartina pectinata Link PI 599561 USA Ingram 19-02 Sporobolus indicus (L.) R. Br. PI 310313 Brazil Ingram 15-99 Stiburus conrathii Hack. PI 11456 South Africa Adams 11456 Tetrachne dregei Nees PI 209829 South Africa Ingram 16-02 Thellungia advena Stapf CANB 468782 Australia Ingram 18-02 Uniola paniculata L. J. I Davis Florida, USA No voucher U. pittieri Hack. Jalisco, Mexico Columbus 4083 (RSA)

identical, but the bootstrap values were generally higher for whether existing infrageneric classifications are consistent the data sets that included gap characters. As they did not with the results of phylogenetic analyses of molecular data. affect the tree topologies, the gap characters were included Recent phylogenetic studies have suggested that Eragros- in the final analyses reported here. tis may need to be split into at least two distantly related The trees from the plastid and nuclear data sets were groups (Van den Borre and Watson 1997; Hilu and Alice largely congruent. This sample of Eragrostis species (with 2001). However, both the plastid and nuclear sequence data the addition of a few segregate genera) is strongly supported in our study support Eragrostis as a monophyletic group as a monophyletic group. This is in striking contrast to re- when a few of the segregate genera are included in it. Genera sults from previous phylogenetic analyses. It seems clear that that almost certainly belong within Eragrostis according to Acamptoclados, Diandrochloa, Neeragrostis, and Pogonar- these data sets include Acamptoclados, Diandrochloa, Neer- thria are embedded within Eragrostis. Thellungia advena, agrostis, and Pogonarthria. This placement of Acamptocla- however, shares a clear affinity with Sporobolus, an associ- dos is not surprising. This species was originally described ation that was highlighted in the original species description as Eragrostis, and the characters used to segregate it were (Stapf 1920). Cladoraphis and Stiburus are found outside weak: both the absence of secondary inflorescence branching Eragrostis in the rps16 data set, but waxy sequences were and free pericarps appear elsewhere in the genus. It was also impossible to obtain, so further data will be necessary before to be expected that Neeragrostis should be returned to Er- they can be accurately placed. agrostis. The major feature distinguishing N. reptans from The simultaneous analysis of the waxy and rps16 sequenc- the rest of Eragrostis is its dioecy, but its other morpholog- es for diploid taxa yielded a single most parsimonious tree (Fig. 3; L ϭ 1750, CI ϭ 56, RI ϭ 59). The infrageneric ical characters clearly ally it with Eragrostis. Its other dis- groups (sensu Clayton and Renvoize 1986; Lazarides 1997; tinguishing feature is its prostrate habit, but N. reptans Cope 1998) to which the various species are assigned are shares this and a number of other morphological features indicated by the letters and numbers adjacent to the taxon with E. hypnoides. These species were resolved as sister taxa names. The cladogram was well resolved, but it is clear that in the rps16 analysis (Fig. 1). the infrageneric groups circumscribed by mode of spikelet Diandrochloa, as represented by D. japonica, is placed disarticulation and other morphological characters do not firmly within Eragrostis by both data sets. Species assigned represent monophyletic groups. Few of these species have to this genus display a number of unique characteristics, in- been anatomically typed, but those that have were assigned cluding a membranous ligule and a distinctive panicle form. to the subgenera of Van den Borre and Watson (1994) and On a gross morphological scale, these differences have mapped on the tree (Fig. 3). According to the limited sample seemed significant enough to some agrostologists to suggest neither subgenus is monophyletic. that these species should be placed in their own genus. How- ever, these molecular data suggest that this genus is in fact DISCUSSION nested within Eragrostis. Further sampling of Diandrochloa species will be necessary to confirm its placement within the Grass taxonomists have struggled with the classification genus. of Eragrostis for many years. Its placement within Chlori- Pogonarthria is firmly placed within Eragrostis in both doideae has been controversial, though recent analyses of several molecular data sets, including both plastid and nu- the plastid and nuclear data sets. This corroborates results clear loci, have all suggested it is sister to Uniolinae (Hilu obtained from phylogenetic analyses of sequence data from and Alice 2001; Columbus et al. 2007), which is consistent the nuclear ITS and plastid trnL–F regions (Columbus et al. with our analysis of waxy (Fig. 2) (relationship unresolved 2007). Morphological characters linking this genus to Er- in our analysis of rps16; Fig. 1). Relationships within the agrostis include three-nerved lemmas and fringed ligules. genus have been difficult to assess. It is a large group with Like Acamptoclados sessilispicus, this genus is characterized an extensive geographic distribution, and other factors such by a lack of secondary inflorescence branching, which has as polyploidy and phenotypic plasticity have confounded at- been thought to be rather important in higher-level classifi- tempts to delimit infrageneric groups. The focus of this study cation in chloridoid grasses. However, the waxy data set has not been on devising a new infrageneric classification, shows that these species are not closely related, suggesting but rather to answer basic questions about whether the genus that this character evolved multiple times within the Er- is monophyletic, how to place its segregate genera, and agrostis clade. At least in this group, a lack of secondary 600 Ingram and Doyle ALISO

inflorescence branching appears not to be an indicator of close relationship. The original description of the monotypic genus Thellun- gia cited an affinity with Sporobolus, which has been confirmed by these analyses. Thellungia advena is placed well away from Eragrostis in a clade with Sporobolus, Spartina Schreb., and Calamovilfa Hack. in both the rps16 and waxy analyses. This result mirrors that found by Ortiz-Diaz and Culham (2000) in an analysis of Sporobolus. These authors found Thellungia advena to be nested well within Sporo- bolus. The placement of Cladoraphis and Stiburus is not yet certain. As mentioned in the Results, waxy sequences could not be obtained for the species in these genera. Additionally, the plastid sequences were of relatively poor quality and included some uncertain base calls. Neither of these taxa is placed within the Eragrostis clade in the rps16 analysis, so they may indeed stand as separate genera, but further data will be necessary to confirm this result. The placement of Sti- burus seems more certain than that of Cladoraphis, however. There are three uncontradicted molecular synapomorphies that support the Stiburus–Uniola L.–Fingerhuthia Nees ex Lehm.–Tetrachne Nees clade, but there are no characters that unequivocally support the placement of Cladoraphis. In general, the molecular data suggest that many of the species that have at times been segregated from Eragrostis based on morphological differences do not actually merit generic status. This is a situation that has been observed in other large genera as molecular data have become increas- ingly important in phylogenetic studies. A well-known ex- ample of this phenomenon is Solanum L. According to D’Arcy (1991), ‘‘Many of the 62 sections [of Solanum] now recognized are of such distinctive appearance that in other plant groups they would be recognized as separate genera.’’ Phylogenetic analyses of molecular data have repeatedly shown that these groups, most notably Lycopersicon Mill., which includes the cultivated tomato and its wild relatives, are best included within a broad Solanum (e.g., Spooner et al. 1993; Bohs and Olmstead 1997; Olmstead and Palmer 1997). A similar situation can be seen in Eragrostis. Many of the species that have been segregated into small genera possess minor but conspicuous morphological differences. To recognize these as genera would require splitting Era- grostis into numerous smaller genera, which would generate both nomenclatural instability and the loss of an easily recognized genus that is generally useful for communication.

Congruence with Previous Cladistic Analyses

The results from the waxy and rps16 analyses are largely congruent with the results from the matK data set collected by Hilu and Alice (2001) when the misidentiﬁed taxa are excluded. Some of the genera are placed differently in their Fig. 1.—Strict consensus of 351 most parsimonious trees (L ϭ analysis, including Dactyloctenium Willd. and Eleusine 230, CI ϭ 72, RI ϭ 88) from the phylogenetic analyses of rps16 sequence data. Bootstrap values are above the branches. Gaertn., but this may be due to their broader sampling within Chloridoideae. However, most of the other major groupings found in our analyses are also present in the matK analysis albeit with some minor rearrangements. These results also agree with the cladograms in Columbus et al. (2007), particularly in terms of the placement of Neeragrostis reptans and Pogonarthria squarrosa. An interesting result from this VOLUME 23 Eragrostis Classiﬁcation 601

Fig. 2.—Strict consensus of three most parsimonious trees (L ϭ 4995, CI ϭ 40, RI ϭ 62) from the phylogenetic analyses of waxy sequence data. Bootstrap values are above the branches. analysis was that Ectrosia leporina R. Br. was ﬁrmly nested phological and anatomical characters shows much less con- within the clade containing all sampled Eragrostis species. gruence with the rps16 and waxy data sets. One of the major Clearly this is another genus that must be considered in fu- problems is that the two subgenera of Eragrostis that the ture analyses of Eragrostis. authors used as terminals in their analyses do not correspond The Van den Borre and Watson (1997) analysis of mor- to any clades found in our analyses of the molecular data, 602 Ingram and Doyle ALISO

Fig. 3.—Single most parsimonious tree (L ϭ 1750, CI ϭ 56, RI ϭ 59) from the simultaneous analysis of rps16 and waxy sequences from diploid species. The columns adjacent to the taxon names show the distribution of various classification systems. Column 1 represents Clayton and Renvoize (1986; E ϭ sect. Eragrostis, L ϭ sect. Lappula, P ϭ sect. Psilantha). Column 2 shows to which group of Lazarides (1997) the taxa belong. Column 3 indicates the groups of Cope (1998). Column 4 represents the subgenera of Van den Borre and Watson (1994; E ϭ subgen. Eragrostis and C ϭ subgen. Caesiae). precluding direct comparisons between the cladograms. of the existing morphology-based classifications to which the However, it is clear that few of the major clades of other taxa we sampled belong are labeled next to the species chloridoid genera recovered in the analyses of our molecular names (Fig. 3). The four sections of Eragrostis recognized data were also found in the Van den Borre and Watson ana- by Clayton and Renvoize (1986) are sect. Platystachya lysis. This suggests that there may be little correspondence Benth., in which the spikelets fall entire; sect. Psilantha (K. between the phylogenetic signals present in the molecular Koch) Tzvelev, which is characterized by the florets disar- data examined so far and the morphological and anatomical ticulating from the top; sect. Eragrostis, whose spikelets dis- characters used in that data set. articulate from the bottom with persistent paleas; and sect. Lappula Stapf, whose spikelets disarticulate from the bottom Congruence with Existing Infrageneric Classifications with the paleas falling with the lemmas. Our taxon sample did not contain any diploid representatives of sect. Platy- The combined analysis of the sequence data for the nu- stachya, but the other three sections are indicated in the first clear and plastid loci from the diploid taxa produced a single column next to the taxon names in Fig. 3. In the Lazarides most parsimonious tree, and the infrageneric groups for three (1997) classification (second column in Fig. 3), Group 5 cor- VOLUME 23 Eragrostis Classification 603 responds to sect. Platystachya, Group 6 corresponds to sect. regions, and a number of unusual Eragrostis species occur Psilantha, and Group 4 corresponds to sect. Lappula. In in these areas that may well belong outside the genus and Group 3 spikelets disarticulate from the bottom with the that were not available for these analyses. Including these rachilla breaking up, while Groups 1 and 2 disarticulate from taxa will be crucial for making informed taxonomic deci- the bottom up with a persistent rachilla. In Group 1 the sions about the delimitation of Eragrostis and its segregates. spikelets are terete or biconvex, whereas they are strongly However, it does appear that some of the large-scale modi- compressed laterally in Group 2. In the Cope (1998) clas- fications to Eragrostis s.s. suggested by the results of some sification (third column in Fig. 3), Group 6 corresponds to previous phylogenetic analyses (Van den Borre and Watson sect. Platystachya, Group 7 is roughly equivalent to sect. 1997; Hilu and Alice 2001) will not be required. Psilantha, Group 9 roughly corresponds to sect. Eragrostis, After determining which species belong in Eragrostis, it and Group 8 roughly corresponds to sect. Lappula. Cope will be important to construct a functional infrageneric clas- also included a number of new groups in his classification. sification, preferably based in part on morphological synapo- Group 1 consists of species with slender panicle branches morphies that can be readily recognized in the field and on that are widely divaricate and have a purple, pilose pulvinus herbarium specimens. Such a resource would provide valu- in each axil. Group 2 corresponds to Cladoraphis and has able clues for identifying plants in this genus, which is a stiff panicle branches that terminate in a naked bristle or task that can be extremely daunting in the geographical re- abortive spikelet. In Group 3, the species have tuberculate- gions where it is species-rich. One of the most serious hin- ciliate palea keels, and Group 4 consists of species with three drances with previous attempts at identifying morphological to five sterile lemmas at the base of the spikelet. In Group characters that may be used to circumscribe infrageneric taxa 5 the lemmas are semi-ovate in profile. The taxon sample has been the inclusion of a large number of allopolyploids was limited for this combined analysis, but it is apparent in the study groups. Allopolyploids are difficult for many that none of these classifications represents monophyletic reasons, most notably the irregular pattern of morphological groups. characters that they exhibit. Development and gene expres- Many of the taxa included in this sample have not been sion are extremely complex and unpredictable in these taxa examined anatomically, but enough data are available to sug- (Wendel 2000), and the aberrant behavior of allopolyploids gest that the classification proposed by Van den Borre and in phylogenetic analyses of morphological data preclude the Watson (1994) may have some value. Subgenus Caesiae use of more rigorous phylogenetic methods to identify mor- (panicoid-type microhairs, PCK-like leaf anatomy) forms a phological synapomorphies for infrageneric taxa. In fact, largely monophyletic group (with the exception of E. trun- only sequence data from low-copy nuclear genes can provide cata), and subgen. Eragrostis (chloridoid-type microhairs, useful information for determining relationships of allopoly- NAD-ME-like leaf anatomy) forms a grade at the base of ploids given that plastid sequences can only elucidate the the tree (Fig. 3). The combination of morphological and ana- history of the maternal progenitor of these taxa. tomical characters suggested by Van den Borre and Watson A possible solution to this problem is to use only diploid (1994) may be more useful than the classifications based on species in reconstructing the evolutionary relationships with- morphology only, but neither subgenus as currently delim- in the genus and to use the resulting cladograms to identify ited is monophyletic, and more data will be necessary to morphological characters that mark the infrageneric groups. fully evaluate this classification. With further refinement and With diploids, it is possible to combine characters from a exploration of these character systems and increased sam- number of sources, including morphology, anatomy, and pling for the molecular data set, these features may be shown both plastid and nuclear sequence data to construct robust to be useful in constructing a functional infrageneric clas- phylogenies. Unfortunately this solution will be difficult to sification. apply in Eragrostis. The proportion of the species in the genus that are thought to be polyploids is large (Hunziker Towards a Classification of Eragrostis and Stebbins 1986), but the ploidy level is not known for all species. For those species that have been subjected to Based on these preliminary analyses of a broad sample of cytological studies, the question of ploidy is not always Eragrostis species and some generic segregates, it appears straightforward. For species where there have been multiple that a number of taxonomic changes will probably be re- chromosome counts, there have been a number of cases quired in this group. The strongest cases for return to Era- where a single morphological species is known to have in- grostis include Acamptoclados and Neeragrostis. Pogonar- dividuals with a range of ploidy levels (e.g., E. curvula has thria should also almost certainly be submerged within an documented chromosome counts ranging from 2n ϭ 40 to expanded Eragrostis. Thellungia, however, does not appear 2n ϭ 80, including euploid and aneuploid numbers; de Wet to be closely related to these taxa and should be removed 1954; Spies and Jonker 1987; Bir and Sahni 1988). A great from Eragrostis. It will be necessary to sample more species deal of cytological work will be necessary to fully under- of Diandrochloa before making any strong recommenda- stand the chromosomal complexity in this group and to fa- tions on its taxonomic status. More data will also be required cilitate the identification of diploids for use in reconstructing for Cladoraphis and Stiburus before it will be possible to evolutionary relationships. determine whether these genera should continue to be re- Even if these difficulties are overcome and a phylogeny cognized. It will also be important in the future to sample of Eragrostis diploids can be used to delimit infrageneric more widely within Eragrostis, with particular emphasis on groups, it will still be extremely complicated to incorporate South African and Australian taxa. Eragrostis and some of the allopolyploids into the taxonomic framework. The results its putative close relatives are particularly diverse in these from the analysis of the waxy data show that several poly- 604 Ingram and Doyle ALISO

ploids derive their homoeologous genomes from widely di- tion. Published by the author, New York, USA. www.cladistics. vergent diploid progenitors (Fig. 2), suggesting that it may org/education.html (Aug 2006). be impossible to assign such taxa to an infrageneric group. HILU, K. W., AND L. A. ALICE. 2001. A phylogeny of Chloridoideae Despite these difficulties, however, this may prove to be the (Poaceae) based on matK sequences. Syst. Bot. 26: 386–405. HUNZIKER, J. H., AND G. L. STEBBINS. 1986. Chromosomal evolution most appropriate strategy for this genus and should provide in the Gramineae, pp. 179–187. In T. R. Soderstrom, K. W. Hilu, a great deal of useful information to guide further research C. S. Campbell, and M. E. Barkworth [eds.], Grass systematics in Eragrostis. and evolution. Smithsonian Institution Press, Washington, D.C., USA. ACKNOWLEDGMENTS INGRAM, A. L., AND J. J. DOYLE. 2003. The origin and evolution of Eragrostis tef (Poaceae) and related polyploids: evidence from The authors thank Travis Columbus, Jerry Davis, Khidir nuclear waxy and plastid rps16. Amer. J. Bot. 90: 116–122. Hilu, Dale Kruse, Carl Lewis, and John Reeder for DNA KOCH, K. 1848. Beitra¨ge zu einer Flora des Orientes. Linnaea 21: and/or plant material; Hailu Tefera, Tiruneh Kefyalew, Ke- 289–443. bebew Assefa, Teshome Dagne, and the staff at the Debre LAZARIDES, M. 1997. A revision of Eragrostis (Eragrostideae, Eleu- Zeit Agriculture Research Center for field support; and Tra- sininae, Poaceae) in Australia. Austral. Syst. Bot. 10: 77–187. vis Columbus for organizing the chloridoid session at Grass- MCDADE, L. 1990. Hybrids and phylogenetic systematics I. Patterns es IV. Thanks also to Travis Columbus, Terry Macfarlane, of character expression in hybrids and their implications for cladistic analysis. Evolution 44: 1685–1700. and Paul Peterson for helpful comments on the manuscript. . 1992. Hybrids and phylogenetic systematics II. The impact This work was supported by NSF grants DEB-0089483 to of hybrids on cladistic analysis. Evolution 46: 1329–1346. JJD and DEB-0098509, the Mario Einaudi Center for Inter- NIXON, K. C. 1999a. WinClada vers. 0.9.99m24. Published by the national Studies, the American Society of Plant Taxonomists author, Ithaca, New York, USA. research grant for graduate students, the Cornell Internation- . 1999b. The parsimony ratchet, a new method for rapid par- al Institute for Food, Agriculture, and Development, the Cor- simony analysis. Cladistics 15: 407–414. nell Institute for African Development, and a National Sci- OLMSTEAD, R. G., AND J. D. PALMER. 1997. Implications for phylo- ence Foundation Graduate Research Fellowship to ALI. geny, classification, and biogeography of Solanum from cpDNA restriction site variation. Syst. Bot. 22: 19–29. ORTIZ-DIAZ, J.-J., AND A. CULHAM. 2000. Phylogenetic relationships LITERATURE CITED of the genus Sporobolus (Poaceae: Eragrostideae) based on nuclear ribosomal DNA ITS sequences, pp. 184–188. In S. W. L. Jacobs AMARASINGHE, V., AND L. WATSON. 1990. Taxonomic significance of and J. Everett [eds.], Grasses: systematics and evolution. CSIRO microhair morphology in the genus Eragrostis Beauv. (Poaceae). Publishing, Collingwood, Australia. Taxon 39: 59–65. PHILLIPS, S. M. 1982. A numerical analysis of the Eragrostideae BIR, S. S., AND M. SAHNI. 1988. Cytomorphological studies on mem- (Gramineae). Kew Bull. 37: 133–162. bers of genus Eragrostis V. Wolf from Punjab Plain (north India). PRENDERGAST, H. D. V., P. W. HATTERSLEY, N. E. STONE, AND M. J. Cytol. Genet. 23: 118–131. LAZARIDES. 1986. C4 acid decarboxylation type in Eragrostis (Po- BOHS, L., AND R. G. OLMSTEAD. 1997. Phylogenetic relationships in aceae): patterns of variation in chloroplast position, ultrastructure, Solanum (Solanaceae) based on ndhF sequences. Syst. Bot. 22: 5– and geographical distribution. Pl. Cell Environm. 9: 333–344. 17. SIMMONS, M. P., AND H. OCHOTERENA. 2000. Gaps as characters in BOR, N. L. 1940. Three new genera of Indian grasses. Indian For- sequence-based phylogenetic analyses. Syst. Biol. 49: 369–381. ester 66: 267–272. SMALL, J. K. 1903. Flora of the southeastern United States. Published BUSH, B. F. 1903. A new genus of grasses. Trans. Acad. Sci. St. by the author, New York, USA. 1370 p. Louis 8: 175–182. SPIES, J. J., AND A. JONKER. 1987. Chromosome studies on African CLAYTON, W. D., AND S. A. RENVOIZE. 1986. Genera graminum: plants. 4. Bothalia 7: 135–136. grasses of the world. Kew Bull., Addit. Ser. 13: 1–389. SPOONER, D. M., G. J. ANDERSON, AND R. K. JANSEN. 1993. Chlo- COLUMBUS, J. T., R. CERROS-TLATILPA, M. S. KINNEY, M. E. SIQUEI- roplast DNA evidence for the interrelationships of tomatoes, po- ROS DELGADO, H. L. BELL, M. P. GRIFFITH, AND N. F. REFULIO- tatoes, and pepinos (Solanaceae). Amer. J. Bot. 80: 676–688. RODRIGUEZ. 2007. Phylogenetics of Chloridoideae (Gramineae): a STAPF, O. 1900. Gramineae, pp. 634–645. In W. T. Thiselton-Dyer preliminary study based on nuclear ribosomal internal transcribed [ed.], Flora Capensis. L. Reeve and Co., Ltd., London, UK. spacer and chloroplast trnL–F sequences, pp. 565–579. In J. T. . 1920. Thellungia, a new genus of Gramineae. Kew Bull. 3: Columbus, E. A. Friar, J. M. Porter, L. M. Prince, and M. G. 96–99. Simpson [eds.], Monocots: comparative biology and evolution— THOMPSON, J. D., T. J. GIBSON, F. PLEWNIAK, F. JEANMOUGIN, AND D. Poales. Rancho Santa Ana Botanic Garden, Claremont, California, G. HIGGINS. 1997. The Clustal࿞X windows interface: flexible strat- USA. egies for multiple sequence alignment aided by quality analysis COPE, T. A. 1998. A synopsis of Eragrostis Wolf (Poaceae) in the tools. Nucl. Acids Res. 25: 4876–4882. Flora Zambesiaca area. Kew Bull. 53: 129–164. VAN DEN BORRE, A., AND L. WATSON. 1994. The infrageneric clas- D’ARCY, W. G. 1991. The Solanaceae since 1976, with a review of sification of Eragrostis (Poaceae). Taxon 43: 383–422. its biogeography, pp. 75–138. In J. G. Hawkes, R. N. Lester, M. , AND . 1997. On the classification of the Chloridoi- Nee, and N. Estrada [eds.], Solanaceae III: taxonomy, chemistry, deae (Poaceae). Austral. Syst. Bot. 10: 491–531. evolution. Royal Botanic Gardens, Kew, Richmond, Surrey, UK. WATSON, L., AND M. J. DALLWITZ. 1992. The grass genera of the DE WET, J. M. J. 1954. Chromosome numbers of a few South Af- world. CAB International, Wallingford, Oxford, UK. 1038 p. rican grasses. Cytologia 19: 97–103. WENDEL, J. F. 2000. Genome evolution in polyploids. Pl. Molec. DE WINTER, B. 1960. A new genus of Gramineae. Bothalia 7: 387– Biol. 42: 225–249. 390. WOLF, N. M. 1776. Genera plantarum vocabulis characteristicis de- GOLOBOFF, P. A. 1993. NONA vers. 2.0. Program and documenta- finita. Ko¨nigsberg, Germany. 177 p.